JPH03255668A - Artificial neural function circuit - Google Patents

Abstract

PURPOSE:To obtain an artificial neural function circuit which is simple in structure, easy to be enhanced in performance and compacted, and similar to an information processing function in a cranial nerve system by a method wherein an organic thin film element which varies in resistance in accordance with an electrical input signal is built in the circuit. CONSTITUTION:Nine organic thin film elements 1-9, three operational amplifiers 10-12, three input terminals 13-15, three output terminals 16-18, and three teacher input terminals 19-21 are provided. When electrical input signals of +10V, +5V, and 0V are given to the input terminals 13, 14, and 15 respectively and an electrical input signal of 0V is given to the teacher input terminals 19-21, a voltage of 0V is applied to the film elements 3, 6, and 9, a voltage of +5V is given to the film elements 2, 5, and 8, and a voltage of 10V is impressed on the film elements 1, 4, and 7. Therefore, the film elements 3, 6, and 9 are kept high in resistance, but the film elements 2, 5, and 8 become low in resistance, and the film elements 1, 4, and 7 become lower in resistance. As the operational amplifiers 10-12 are all equal in input signal, the outputs of the output terminals 16-18 are equal to each other.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application This invention relates to a pseudo-neural function circuit useful in the electronic industry field, particularly in the information processing industry field.

Conventional technology In recent years, there has been active research into new information processing devices that imitate the information processing functions (neural functions) of the brain and nerves, so-called neurocomputers, neural networks, etc., including character recognition systems. It has come to be applied to various systems. These new information processing functions are realized by using existing inorganic semiconductor devices, impedance circuits, etc., on hardware or software.

For example, Hopfield used an impedance circuit composed of an amplifier, a resistor button, and a capacitor to show that the Ei and Lugi functions of this neural network converge to take a minimum value. , it was shown that combinatorial problems can be solved. In addition, it has been proven to be effective for problems such as associative memory and pattern recognition, and is attracting attention as a new, unprecedented information processing function.

Problems to be Solved by the Invention However, even if this new information processing function is realized on software or hardware,
All of these devices are composed of existing semiconductor devices or combinations of these devices, so their configurations are complex and impractical. Moreover, there are limits to its functionality and miniaturization of elements.

In view of these circumstances, the present invention provides a pseudo-neural function circuit similar to the information processing function of the nervous system of the brain, which has a simple configuration and is easy to improve functionality and downsize. That is the issue.

Means for Solving the Problems In order to solve the above problems, the pseudo-neural functional circuit according to the first aspect incorporates an organic thin film element whose resistance state changes according to an electrical input signal.

Specifically, this organic thin film element maintains the changed resistance state for at least a certain period of time after the electrical input signal is released, as in the invention of claim 2, or maintains the changed resistance state for at least a certain period of time as in the invention of claim 3. It is a thin film element that has the physical property that the direction of change (increase/decrease direction) of the electric current depends on the polarity of the electrical input signal.

An example of a thin film having such physical properties is a lead phthalocyanine thin film, but the present invention is not limited thereto.

Function: Since the pseudo neural function circuit according to the present invention uses an organic thin film element, it has a simpler configuration and is more practical than when using existing semiconductor devices. In addition, since organic thin film elements can be easily provided with various characteristics, they can be made more sophisticated in functionality and more compact (device miniaturization).

In particular, lead phthalocyanine thin films retain the changed resistance state for at least a period of time even after the electrical input signal is removed, and furthermore, the direction of change (increase/decrease direction) of the resistance state depends on the polarity of the electrical input signal. In this case, it is easier to improve functionality and make it more compact.

EXAMPLES Examples of the present invention will be described below, but before that, the electrical characteristics of the organic thin film element itself will be described as follows: An element with a lower electrode on the back surface and an upper electrode on the front surface of a lead phthalocyanine vacuum-deposited thin film with a thickness of 1.2 μm. will be explained using an example.

The manner in which the resistance of the thin film changes when the upper electrode is grounded and a positive/negative sweep voltage (electrical input signal) of 4 V/sec is applied to the lower electrode will be explained with reference to FIG.

In this case, the lead phthalocyanine thin film is initially in a high resistance state R1, but when the sweep voltage exceeds the positive voltage 1, it changes to a low resistance state R2 through a negative resistance region. Thereafter, when the sweep voltage exceeds the negative voltage (2), it passes through the negative resistance region and returns to the initial high resistance state R1. In other words, one positive and negative voltage sweep causes the circuit to reciprocate between the high resistance state R1 and the low resistance state R2.

After the positive voltage exceeds ■1 and changes to low resistance state R2, if the electrical input signal is released without moving to negative voltage sweep, low resistance state R
2 is retained. The retention is not permanent, and the high resistance state R1 is naturally restored after a certain period of time. Of course, it is also possible to use a thin film that does not naturally return to the high resistance state R1 even after a certain period of time has passed, but maintains the low resistance state R2 all the time.

After changing to low resistance state R2, it does not shift to negative voltage sweep,
When the positive voltage sweep is performed again, the voltage changes according to the voltage-current characteristics shown in the graph of FIG. 3, and no change occurs in maintaining the low resistance state R2.

Conversely, no matter how many negative voltage sweeps are performed in the initial state of high resistance state R1, the voltage shown in the graph of FIG.
The high resistance state R1 is maintained only by changing according to the current characteristics, and no change to the low resistance state R2 occurs. These results show that the direction of change (increase or decrease) in the resistance state of the lead phthalocyanine thin film depends on the polarity of the applied voltage (electrical input signal).

Next, we will discuss the relationship between voltage value and temporal response when a constant voltage is applied to the lead phthalocyanine thin film.

As shown in Figure 5, the change in resistance value, that is, the speed of change from a high resistance state to a low resistance state, depends on the applied voltage (hp), and the higher the voltage, the faster the change from a high resistance state to a low resistance state. The time to do it becomes faster. Furthermore, the resistance value in the low resistance state reached is a value that depends on the applied voltage. If the application time is the same, the higher the applied voltage, the lower the attained resistance value. Of course, each low resistance state reached is maintained for at least a certain period of time even after the voltage application is removed.

As can be seen from this, the lead phthalocyanine thin film is
It is also possible to store an intermediate resistance state between the high resistance state R1 and the low resistance state R2, and it also becomes a multi-value memory element.

FIG. 6 shows voltage-current characteristics of the lead phthalocyanine thin film that are slightly different from those in FIG. 2. In FIG. 6, the voltage rises sharply in the positive voltage region and is asymmetrical when viewed from the zero point. Therefore, during a positive voltage sweep, the button changes to a low resistance state immediately after voltage is applied, but even after entering a negative voltage sweep, it takes time to return to a high resistance state.

That is, in the case of FIG. 6, the resistance change speed of the lead phthalocyanine thin film is fast from a high resistance state to a low resistance state, and conversely, it is slow from a low resistance state to a high resistance state.

In this way, organic thin film devices using lead phthalocyanine thin films exhibit various resistance changes and can easily exhibit advanced functions. For example, one of the important functions of neurons is the plasticity of synaptic connections. This is because the strength of synaptic connections changes depending on the strength and number of stimulation inputs, and this makes information processing possible in living organisms. By using , it is possible to easily realize a circuit that operates in a sufficiently similar manner.

This invention is not limited to the above-mentioned examples. For example, the organic thin film may be made of materials other than lead phthalocyanine. Alternatively, the electrical input signal may be provided as a current.

FIG. 1 shows an embodiment of the pseudo neural function circuit of the present invention.

This pseudo neural functional circuit consists of nine organic thin film elements 1 to 9.
Three operational amplifiers 10-12. Three input terminals 13-1
5. It is equipped with three output terminals 16-18, a button, and three teacher input terminals 19-21. Note that 22 to 24 are fixed resistance elements.

One side of the organic thin film elements 1 to 3 is connected to the input terminals 13 to 15, respectively, and the other side is connected to the inverting input terminal (
(negative terminal). Organic thin film elements 4 to 6
One side is connected to the input terminals 13 to 15, respectively, and the other side is connected to the inverting input terminal of the operational amplifier 11. One side of the organic thin film elements 7 to 9 is connected to the input terminals 13 to 15, respectively, and the other side is connected to the inverting input terminal of the operational amplifier 12. The teacher input terminals 19-21 are connected to non-inverting input terminals (plus terminals) of operational amplifiers 10-12, respectively.

Organic thin film devices 1 to 9 use lead phthalocyanine thin films, and as described above, are devices having various characteristics shown in FIGS. 2 to 5. Each element 1 to 9 has a gold vapor-deposited film for a lower electrode formed on the surface of an insulating substrate, and a lead phthalocyanine thin film (organic thin film) with a film thickness of 1.2 μm is laminated thereon by a vacuum vapor deposition method. It is made by forming a gold vapor-deposited film for the upper electrode.

Next, the operation of the pseudo neural function circuit shown in FIG. 1 will be explained.

1. +IOV to input terminal 13, +5V to input terminal 14
, an OV electrical input signal is given to the input terminal 15, and an OV electrical input signal is given to the teacher inputs 19-21.

In this case, the voltage applied to the organic thin film element is
is 0■, membrane 2.5.8 is +5V, membrane 1, 4.7 is +to
It becomes V. Therefore, organic thin film element 3.6.9 is in a high resistance state, organic thin film element 2.5.8 is in a low resistance state, and organic thin film elements 1 and 4.7 are in a lower resistance state. becomes. Since the input signal amount of each operational amplifier 10-12 is equal, the same amount of output is outputted to the output terminals 16-18.

Next, set the teacher power 19 to 110V, and set the teacher input 2 to 110V.
If you apply +20V to 0 and +20V to teacher power 21,
Since a sufficient negative voltage is applied to the organic thin film elements 4 to 9, they return to a high resistance state. Of course, organic thin film elements 1 to 3 to which no negative voltage is applied are 11 in a low resistance state. Therefore, only the output of the output terminal 16 is maintained at its full value, and the outputs of the output terminals 17 and 18 disappear.

Even if the voltage applied to the input terminal and teacher input terminal is cut off, the resistance state of each organic thin film element does not change. Therefore, when the same voltage is applied to the input terminals 13 to 15 next time, only the output terminal 16 is outputted.

By changing the pattern of voltage application to the teacher input terminals 19 to 21, it is also possible to output output to the output terminals 17 and 18 in the same way. Therefore, a difference in the input pattern can be detected by the output/non-output mode of the button on the output terminal.

Next, the pseudo neural circuit shown in Fig. 1 was constructed in exactly the same manner except that the characteristics of the organic thin film element were changed from those shown in Fig. 2 to those shown in Fig. 6. Then, as before, the input voltage is +IOV at input terminal 13 and +5■ at input terminal 14.
, the input terminal 15 is multiplied by OV. On the other hand, the teacher input is OV to input terminal 19, +5■ to input terminal 20,
+5■ was also applied to the input terminal 21. In this case, only the thin film 1 is in the lowest resistance state and the output from the 1 output terminal 16 is the highest. After this, even if the input and teacher input are turned off, this state will be maintained, so next input terminals 13~
If the same input is applied to terminal 15, a large output will be output only from output terminal 16. Furthermore, since the speed of change from a low resistance state to a high resistance state is slow and the speed of change from a high resistance state to a low resistance state is fast, learning can be completed in a short time and is difficult to forget. By changing the input pattern, it is possible to output the signals from the output terminals 17 and 18 in the same way, so that in this case as well, the difference in the input pattern can be detected by the output.

In FIG. 7, using input terminals 13 and 14, (+lOV
, OV) and (QV, +10V) are shown to show the output results of the output terminals 16 and 17 when learning the two patterns. ○ indicates output from output terminal 16, ・ indicates output terminal 17
Shows the output from As you can see from the figure, learning
For each pattern (-5■, -2■), (-
2■, -5■), it can be seen that the two patterns can now be distinguished.

This means that a persebutron, a type of neural network, has been realized in hardware by utilizing the physical properties of materials.

Any specific input pattern may be used. For example, if the input pattern is created using signals from various sensors, it can be applied to various recognition devices.

In this embodiment, the number of inputs is three, but it may be increased to four or more to provide a -layer or advanced function.

Effects of the Invention As stated above, the pseudo-neural functional circuits of claims 1 to 4 incorporate organic thin film elements, so they have a simple configuration and are easy to implement for practical use in terms of functional sophistication and compactness. Furthermore, in the pseudo neural function circuit of claim 2, the organic thin film element maintains the changed resistance state for at least a certain period of time even after the electrical input signal is released. In the pseudo-neural functional circuit of claim 3, the direction of change (increase/decrease direction) of the resistance state depends on the polarity of the electrical input signal, and in the pseudo-neural functional circuit of claim 4, the speed of change of the resistance value is decreased compared to when it increases. By speeding up time, it is easier to achieve higher functionality.

[Brief explanation of drawings]

FIG. 1 is a pseudo neural function circuit diagram of an embodiment of the present invention;
Figures 2 to 4 show the voltage-current characteristics of the organic thin film element used in the pseudo-neural function circuit of this invention, and Figure 5 shows the voltage-current characteristics that flow when a constant voltage is applied to this organic thin film. Figure 6 is a characteristic diagram showing the time change of current.
FIG. 7 is a diagram showing output characteristics when two patterns are discriminated using the pseudo neural function circuit according to the present invention. 1-9... Organic thin film element, 10-12... Operational amplifier, 13-15... Input terminal, 16-18... Output terminal,
19-21...Teacher input terminals.

Claims (4)

[Claims] (1) A pseudo-neural functional circuit incorporating an organic thin film element whose resistance state changes depending on an electrical input signal. (2) The pseudo neural function circuit according to claim 1, wherein the organic thin film element maintains the changed resistance state for at least a certain period of time after the electrical input signal is released. (3) The pseudo neural function circuit according to claim 1 or 2, wherein the direction of change in the resistance state depends on the polarity of the electrical input signal. (4) The pseudo neural function circuit according to any one of claims 1 to 3, wherein the rate of change in the resistance value of the organic thin film is faster when the resistance decreases than when the resistance increases.